Quantifying Associations between Environmental Stressors and Demographic Factors

摘要

Association rule mining (ARM) [1, 2], also known as frequent item set mining [3] or market basket analysis [2], has been widely applied in many different areas, such as business product portfolio planning, intrusion detection infrastructure design, gene expression analysis, medical diagnosis, and drug prescription pattern. In recent years, ARM has also been used to analyze relationships between environmental stressors and adverse human health effects [4, 5]. We employed ARM to identify and quantify associations within and between ambient pollutants (environmental stressors) and demographic factors such as age, poverty, race/ethnicity, and education attainment. Specifically, we linked the 2011 NATA (National-Scale Air Toxics Assessment) U.S. Census tract-level air pollutant exposure concentration data with the 2010-2014, 5-Year Summary Files in the American Community Survey (ACS), and created relevant chemical and demographic variables. Association rules were generated based on the merged data (NATA Data and ACS 5-Year Summary Files) and filtered with specific criteria or measurements to enhance understanding of the relationships between multiple chemical stressors and socio-demographic factors. We also utilized a graph-based visualization tool [6] to depict the interacting relations among all the stressors or factors that play active roles in the resultant rules. Our main aim is to demonstrate the ability of using unsupervised data mining methods to identify associations among multiple stressors (e.g., to find the underlying structure of and the relationship[s] between the stressors), which can be useful for assessment of co-exposure to chemical and non-chemical stressors, and informative for public health decision-making, especially when it comes to addressing environmental justice issues and social disparities.